Project Summary Acute myeloid leukemia (AML) is a genetically heterogeneous disease, which occurs de novo or evolves from pre- existing myeloid disorders. The incidence of AML increases with age, and with increasing age, prognosis and overall survival decline. Treatment failure in many patients is attributed to relapsed disease and remains the fundamental clinical challenge in these patients. The exact biological basis of relapse remains unclear. One prevailing theory is that inherently chemotherapy resistant leukemia stem cells (LSCs) survive and repopulate the disease. However genetic background must also have an influence on LSC functions since specific mutations are associated with more frequently relapsing disease. It has also been observed that AMLs undergo clonal evolution with relapse, suggesting that pre-existing subclones or chemotherapy induced natural selection contribute to this phenomenon. Our preliminary data strongly suggest that epigenetic mechanisms play a fundamental role in this process. Specifically: i) we observe extensive epigenetic deregulation in all AML patients, ii) specific epigenetic signatures and classifiers are associated with poor outcomes, iii) there is a relative paucity of genetic mutations in AML as compared to solid tumors (based on TCGA data), iv) many of the common genetic lesions in AML directly alter epigenetic gene regulation. Moreover in preliminary studies we observe that AMLs at relapse display large scale epigenetic changes, whereas TCGA finds relatively subtle genetic changes at relapse. Of note, many of the epigenetic changes we observe are in common between patients, suggesting that AMLs may respond and resist chemotherapy using shared or stereotyped mechanisms. Based on these and other considerations we hypothesize that AML relapse and adaptation to chemotherapy is encoded by a combination of epigenetic and genetic lesions that are either present in a subset of AML cells at diagnosis or occur de novo in response to exposure to cytotoxic drugs. We predict that these lesions preferentially disrupt specific biological pathways that directly mediate chemotherapy resistance, and that detection of these lesions at diagnosis will be predictive of unfavorable outcome. In order to test these hypotheses we will perform a genome wide integrative genetic and epigenetic comparison of seventy matched diagnosis/relapse AML specimens including the use of novel methods developed by the applicant. Results will be validated, and the clinical significance determined in an independent cohort of 750 patients all enrolled in a single phase II clinical trial. The biological function of relapse-associated lesions will be confirmed in functional assays, with the ultimate goal of developing targeted therapies to prevent leukemia relapse.